Some recent discoveries in the ChatGPT web app and Android beta versions hint at a powerful new feature—something similar to OpenAI’s existing “Operator” tool. As noticed by a user named Tibor on X (formerly Twitter), the Android app’s code now includes phrases like “click,” “drag,” “type,” and even “terminal feed.” These clues suggest that ChatGPT might soon gain the ability to interact with a remote browser or a secure environment to perform tasks for users.
For context, OpenAI already has a tool called Operator, which lets an AI agent control a remote browser to carry out actions on your behalf—things like navigating websites or running tasks online.
The new code references also mention ChatGPT checking available APIs and reading API documentation, implying it could soon use external tools or services to complete tasks. Plus, there are mentions of “computer tool” functions that involve clicking and performing actions on a computer, pointing to broader system-level control.
Interestingly, the code also refers to an “intake form,” hinting that OpenAI may initially limit access to this feature through an invite-only beta program before making it available to everyone.
Could this be part of ChatGPT-5 or another model? We can’t say for sure. And with OpenAI currently focused on its rivalry with Meta, it may be a while before we learn more.
0 Comments